Patients with pulmonary infiltrates receive more than half of the antibiotics prescribed in the ICU. Many of these patients receive antibiotics for suspected, but not proven, respiratory infections. Yet many of these patients with abnormal radiographs actually have a non-infectious etiology. Although pneumonia is the most common hospital acquired infection in ICU patients, it is difficult to diagnose definitively because the diagnosis relies heavily on radiographic findings that correlate poorly with clinical pneumonia. But because of the attributable mortality associated with nosocomial pneumonia, the risk of missing a treatable infection often outweighs the perception of minimal risk antibiotic therapy. Consequently, overuse or inappropriate use of antibiotics for the treatment of suspected pneumonia is widespread in this subset of patients, and the emergence of antimicrobial resistant bacteria as a result of antibiotic overuse is increasingly common. This protocol is designed to determine whether short course empiric antibiotic therapy (3 days of meropenem) for patients with new pulmonary infiltrates can reduce the emergence of antimicrobial resistant organisms compared to a standard course of antibiotic therapy (10 days of therapy with antibiotics of the primary care teams choosing). In addition, it will determine whether hospital length of stay can be reduced, hospital cost can be reduced, while morbidity and mortality of patients will not be affected. This protocol targets patients who have new pulmonary infiltrates, yet are at low risk of having pneumonia, as determined using the Clinical Pulmonary Infection score (CPIS). Progress Report: Final approval from the BAMSG Steering Committee has been obtained. The study has undergone scientific review and is pending IRB review at the Clinical Center once ten study sites are identified and approved. Site surveys are being circulated to assess the sites ability to enroll patients.